On January 29th David Rose wrote an article in the Mail on Sunday showing a graph of annual average global temperatures since 1997 according to the most commonly used version of the HadCrut3 database.
Fig 1. Click on image to enlarge.
It showed no increase in temperature. It has been often stated that there has been no statistically significant warming in the various temperature datasets since the start of this century, and this is now generally accepted. Pushing it back a few more years is possible. The years 1999 and 2000 were slightly cooler than the post-2001 data but since 1998, a strong El Nino year, is higher they provide compensation for each other. Thus if you look at the trend since 1997 you can easily see, and calculate, that it is negligible, (a simple linear trendline over the period is about 0.015 deg C with large errors. For comparison Noaa data has 0.07 and NasaGiss 0.16 over the same period, both small.)
It is also possible to push it back a few more years still, according to the brief calculations made by Prof Phil Jones. It seems that 1995 – 2009 is flat, but 1995 – 2010 has a slight positive trend, though not at any impressive significance. It should be noted that 1995 – 2011 is back to no significant increase.
Statistically speaking it is accurate to say that according to HadCrut3 the world’s temperature has not increased for the 16 years between 1995 and 2011, though many prefer the more conservative ten years post-2001. This is not a ‘sceptical’ claim just a straightforward description of the data.
The UK Met Office didn’t like the Mail on Sunday article and issued a repost on its blog, saying that there is no standstill as the average global temperatures for the decades of the 80s, 90s, and 00s show a rising trend. Hence, they argue, global warming has not slowed down.
Fig 2. Click on the image to enlarge.
The relevant figures are easily calculated from HadCrut3.
|Average Temperature Anomaly|
|1981 – 1990||0.10|
|1991 – 2000||0.24|
|2001 – 2010||0.43|
The Met Office said, “…for Mr Rose to suggest that the latest global temperatures available show no global warming in the last 15 years is entirely misleading…what is absolutely clear is that we have continued to see a trend of warming with the decade of 2000-2009 being clearly the warmest in the instrumental record going back to 1850.”
However as they say, it’s not what you say but what you fail to say. The decadal graph on its own misrepresents the situation. The thing the Met Office omitted to say in their “refutation” is that their graph and the one used in the Mail on Sunday are perfectly consistent with each other. Indeed they both come from the very same Met Office data! They are two different ways of accurately describing the same data.
The world has warmed since the start of the current warming spell that started around 1980 and each decade has been warmer than the previous one. But that is not all. We are very used to looking an annual temperatures and decadal averages, but we should be aware of what averaging does, what it emphasises and what it obscures.
There is no particular reason to choose to average over a decade. It is an example of cherry-picking as whilst nature knows about annual cycles the climate knows nothing about how many fingers we have or that it is the arithmetical base of our counting system. Neither does it know which year is designated as 1980, 1990 etc. One could equally well choose 5, 9, 12, 15 years for example, and if you do you will see different things in the data. Decadal averages are convenient, but if you use them blindly you destroy valuable information in the dataset. The world does not react in multiples of ten years.
Looking at the three decades with annual HadCrut3 data it is apparent that there is not much of an increase in the 80s, more of an increase in the 90s and no increase in the 00s. The lack of increase since 1997 is apparent despite the Met Office saying that “no warming in the last 15 years is entirely misleading.” From 1997 this is exactly the Mail on Sunday’s graph.
Fig 3. Click on image to enlarge. (Data is for 1980 -2011. Error bars are set at 0.1 deg C)
A more revealing and sophisticated way of analysing the data is to look at the data and note its structure. The fact that the temperature standstill extends from the present time back to 1997 obviously indicates that what is an important aspect of the data is being spread into two decadal databins.
Yearly data and decadal data are often used for analysis. Because of the integration noise is reduced in decadal data when compared to annual data, but another way is to look at five-year means which will have intermediate characteristics, and show some degree of the structure that is wiped out in decadal integrations.
|Average Temperature Anomaly|
|1971 – 1975||-0.08|
|1976 – 1980||-0.05|
|1981 – 1985||0.05|
|1986 – 1990||0.15|
|1991 – 1995||0.16|
|1996 – 2000||0.32|
|2001 – 2005||0.45|
|2006 – 2010||0.41|
Five Year Averages. Source HadCrut3.
As we have mentioned before there is another version of HadCrut3 (calculated a slightly different way) which we should show for completeness. Essentially it shows the same thing (if it didn’t there would be big implications) so I prefer to stick to the most quoted version.
Fig 4. Post-1970 using the ‘other’ HadCrut3. Click on image to enlarge.
Fig 5. As above but post 1997.
Fig 5 produces a simple linear trendline of about 0.05 deg C with large errors.
Looking at these figures, in conjunction with Figure 3 shows some interesting aspects of the data, when you consider the errors:
The second 5 years of the 70s were similar to the first 5 years of the 80s and the first 5 years of the 70s were not much different from them (0.05, -0.05 and -0.08 respectively).
The second 5 years of the 80s were warmer than the first 5 years of the 80s and similar to the first 5 years of the 90s (0.15 and 0.16).
The first 5 years of the 90s had a temperature anomaly that was half of that in the second 5 years of the 90s (0.16 and 0.32).
The second 5 years of the 90s and both 5-year periods of the 00s are comparable (0.32, 0.45, 0.41).
The second 5-year period of the 00s was cooler than the first (0.41 and 0.45).
What five-year data does is to show that of the 30 years of the recent warm spell the second half shows little change. There is little change between 1986 -1995, and there is also little change in the first decade. You can see this effect if you plot trends from a specific start date and go back or forward each year through the data. This approach supports the above conclusions and could be interpreted as saying that global warming has ‘stopped’ many times since 1980. Such an interpretation is however, too crude, and one should stand back from trends drawn between consecutive years and look at the wider picture, which is that the rate of global warming as inferred from the data of the past 30 years is variable. Such a conclusion is rather obvious since the post 1980 data is not well represented by a straight line!
What this means is that the data shows that the only significant warming phase of the current warming period was in the middle of the 90s leading up to the El Nino which took temperatures higher, at which they stayed constant. Such a scenario tells a very different story from the Met Office’s graph that implies consistent sustained warming over three decades. It is accurate, but it is partial and crude. To parade it alone as proof of no standstill and consistent warming is misleading.
Another curiosity is the fall in temperature seen over the past decade. If you just consider the past ten years the fall in Hadcrut3 is:
|Average Temperature Anomaly.|
|2002 – 2006||0.45|
|2007 – 2011||0.39|
It is also apparent in NasaGiss, though barely (consider the comments in the NasaGiss 2011 summary, “The 5-year (60-month) running mean global temperature hints at a slowdown in the global warming rate during the past few years.”) (Though if you consider 2001-05 and 2006 -10 you get 54 and 55.4, so caution is urged).
|Average Temperature Anomaly|
|2007 – 2011||54.8|
Personally when errors are considered I prefer a more conservative assessment of the post-1997 data consistent with a straight line of zero gradient. However in the analysis above I have just considered means, for comparison with what seems normal practice with those who analyse the global temperature databases.
The data shows that global warming is not consistent. It is an isolated event in-between periods of stable temperature. How this relates to GHG forcing is another matter.
So whilst both graphs, the Mail on Sunday and the Met Offices are correct, it is the Mail on Sunday’s that is the more fundamental, as it is less massaged.
It is curious that the small data set of the recent warming spell (only thirty numbers or so if annual data is used) has been argued about, analysed and misrepresented for so long by climatologists and activists. In most other disciplines I suggest its simple properties would have been extracted much more simply minus the various “interpretations.”
In the NasaGiss summary of 2011 James Hansen, Reto Ruedy, Makiko Sato and Ken Lo, say
“Thus, although the current global warming graphs are suggestive of a slowdown in global warming, this apparent slowdown may largely disappear as a few more years of data are added…We conclude that the slowdown of warming is likely to prove illusory, with more rapid warming appearing over the next few years.”
The purpose of my quoting this Nasa press release is the use of words. The slowdown is ‘apparent’ and will prove ‘illusory’ if warming resumes.
The slowdown/standstill is not ‘apparent,’ it is real. In addition, if warming resumes the standstill will not prove ‘illusory,’ but rather temporary. This is not the way to dispassionately describe scientific data.